1/* 2 * Generic EDAC defs 3 * 4 * Author: Dave Jiang <djiang@mvista.com> 5 * 6 * 2006-2008 (c) MontaVista Software, Inc. This file is licensed under 7 * the terms of the GNU General Public License version 2. This program 8 * is licensed "as is" without any warranty of any kind, whether express 9 * or implied. 10 * 11 */ 12#ifndef _LINUX_EDAC_H_ 13#define _LINUX_EDAC_H_ 14 15#include <linux/atomic.h> 16#include <linux/device.h> 17#include <linux/completion.h> 18#include <linux/workqueue.h> 19#include <linux/debugfs.h> 20 21struct device; 22 23#define EDAC_OPSTATE_INVAL -1 24#define EDAC_OPSTATE_POLL 0 25#define EDAC_OPSTATE_NMI 1 26#define EDAC_OPSTATE_INT 2 27 28extern int edac_op_state; 29extern int edac_err_assert; 30extern atomic_t edac_handlers; 31extern struct bus_type edac_subsys; 32 33extern int edac_handler_set(void); 34extern void edac_atomic_assert_error(void); 35extern struct bus_type *edac_get_sysfs_subsys(void); 36extern void edac_put_sysfs_subsys(void); 37 38enum { 39 EDAC_REPORTING_ENABLED, 40 EDAC_REPORTING_DISABLED, 41 EDAC_REPORTING_FORCE 42}; 43 44extern int edac_report_status; 45#ifdef CONFIG_EDAC 46static inline int get_edac_report_status(void) 47{ 48 return edac_report_status; 49} 50 51static inline void set_edac_report_status(int new) 52{ 53 edac_report_status = new; 54} 55#else 56static inline int get_edac_report_status(void) 57{ 58 return EDAC_REPORTING_DISABLED; 59} 60 61static inline void set_edac_report_status(int new) 62{ 63} 64#endif 65 66static inline void opstate_init(void) 67{ 68 switch (edac_op_state) { 69 case EDAC_OPSTATE_POLL: 70 case EDAC_OPSTATE_NMI: 71 break; 72 default: 73 edac_op_state = EDAC_OPSTATE_POLL; 74 } 75 return; 76} 77 78/* Max length of a DIMM label*/ 79#define EDAC_MC_LABEL_LEN 31 80 81/* Maximum size of the location string */ 82#define LOCATION_SIZE 256 83 84/* Defines the maximum number of labels that can be reported */ 85#define EDAC_MAX_LABELS 8 86 87/* String used to join two or more labels */ 88#define OTHER_LABEL " or " 89 90/** 91 * enum dev_type - describe the type of memory DRAM chips used at the stick 92 * @DEV_UNKNOWN: Can't be determined, or MC doesn't support detect it 93 * @DEV_X1: 1 bit for data 94 * @DEV_X2: 2 bits for data 95 * @DEV_X4: 4 bits for data 96 * @DEV_X8: 8 bits for data 97 * @DEV_X16: 16 bits for data 98 * @DEV_X32: 32 bits for data 99 * @DEV_X64: 64 bits for data 100 * 101 * Typical values are x4 and x8. 102 */ 103enum dev_type { 104 DEV_UNKNOWN = 0, 105 DEV_X1, 106 DEV_X2, 107 DEV_X4, 108 DEV_X8, 109 DEV_X16, 110 DEV_X32, /* Do these parts exist? */ 111 DEV_X64 /* Do these parts exist? */ 112}; 113 114#define DEV_FLAG_UNKNOWN BIT(DEV_UNKNOWN) 115#define DEV_FLAG_X1 BIT(DEV_X1) 116#define DEV_FLAG_X2 BIT(DEV_X2) 117#define DEV_FLAG_X4 BIT(DEV_X4) 118#define DEV_FLAG_X8 BIT(DEV_X8) 119#define DEV_FLAG_X16 BIT(DEV_X16) 120#define DEV_FLAG_X32 BIT(DEV_X32) 121#define DEV_FLAG_X64 BIT(DEV_X64) 122 123/** 124 * enum hw_event_mc_err_type - type of the detected error 125 * 126 * @HW_EVENT_ERR_CORRECTED: Corrected Error - Indicates that an ECC 127 * corrected error was detected 128 * @HW_EVENT_ERR_UNCORRECTED: Uncorrected Error - Indicates an error that 129 * can't be corrected by ECC, but it is not 130 * fatal (maybe it is on an unused memory area, 131 * or the memory controller could recover from 132 * it for example, by re-trying the operation). 133 * @HW_EVENT_ERR_FATAL: Fatal Error - Uncorrected error that could not 134 * be recovered. 135 */ 136enum hw_event_mc_err_type { 137 HW_EVENT_ERR_CORRECTED, 138 HW_EVENT_ERR_UNCORRECTED, 139 HW_EVENT_ERR_FATAL, 140 HW_EVENT_ERR_INFO, 141}; 142 143static inline char *mc_event_error_type(const unsigned int err_type) 144{ 145 switch (err_type) { 146 case HW_EVENT_ERR_CORRECTED: 147 return "Corrected"; 148 case HW_EVENT_ERR_UNCORRECTED: 149 return "Uncorrected"; 150 case HW_EVENT_ERR_FATAL: 151 return "Fatal"; 152 default: 153 case HW_EVENT_ERR_INFO: 154 return "Info"; 155 } 156} 157 158/** 159 * enum mem_type - memory types. For a more detailed reference, please see 160 * http://en.wikipedia.org/wiki/DRAM 161 * 162 * @MEM_EMPTY Empty csrow 163 * @MEM_RESERVED: Reserved csrow type 164 * @MEM_UNKNOWN: Unknown csrow type 165 * @MEM_FPM: FPM - Fast Page Mode, used on systems up to 1995. 166 * @MEM_EDO: EDO - Extended data out, used on systems up to 1998. 167 * @MEM_BEDO: BEDO - Burst Extended data out, an EDO variant. 168 * @MEM_SDR: SDR - Single data rate SDRAM 169 * http://en.wikipedia.org/wiki/Synchronous_dynamic_random-access_memory 170 * They use 3 pins for chip select: Pins 0 and 2 are 171 * for rank 0; pins 1 and 3 are for rank 1, if the memory 172 * is dual-rank. 173 * @MEM_RDR: Registered SDR SDRAM 174 * @MEM_DDR: Double data rate SDRAM 175 * http://en.wikipedia.org/wiki/DDR_SDRAM 176 * @MEM_RDDR: Registered Double data rate SDRAM 177 * This is a variant of the DDR memories. 178 * A registered memory has a buffer inside it, hiding 179 * part of the memory details to the memory controller. 180 * @MEM_RMBS: Rambus DRAM, used on a few Pentium III/IV controllers. 181 * @MEM_DDR2: DDR2 RAM, as described at JEDEC JESD79-2F. 182 * Those memories are labed as "PC2-" instead of "PC" to 183 * differenciate from DDR. 184 * @MEM_FB_DDR2: Fully-Buffered DDR2, as described at JEDEC Std No. 205 185 * and JESD206. 186 * Those memories are accessed per DIMM slot, and not by 187 * a chip select signal. 188 * @MEM_RDDR2: Registered DDR2 RAM 189 * This is a variant of the DDR2 memories. 190 * @MEM_XDR: Rambus XDR 191 * It is an evolution of the original RAMBUS memories, 192 * created to compete with DDR2. Weren't used on any 193 * x86 arch, but cell_edac PPC memory controller uses it. 194 * @MEM_DDR3: DDR3 RAM 195 * @MEM_RDDR3: Registered DDR3 RAM 196 * This is a variant of the DDR3 memories. 197 * @MEM_DDR4: DDR4 RAM 198 * @MEM_RDDR4: Registered DDR4 RAM 199 * This is a variant of the DDR4 memories. 200 */ 201enum mem_type { 202 MEM_EMPTY = 0, 203 MEM_RESERVED, 204 MEM_UNKNOWN, 205 MEM_FPM, 206 MEM_EDO, 207 MEM_BEDO, 208 MEM_SDR, 209 MEM_RDR, 210 MEM_DDR, 211 MEM_RDDR, 212 MEM_RMBS, 213 MEM_DDR2, 214 MEM_FB_DDR2, 215 MEM_RDDR2, 216 MEM_XDR, 217 MEM_DDR3, 218 MEM_RDDR3, 219 MEM_DDR4, 220 MEM_RDDR4, 221}; 222 223#define MEM_FLAG_EMPTY BIT(MEM_EMPTY) 224#define MEM_FLAG_RESERVED BIT(MEM_RESERVED) 225#define MEM_FLAG_UNKNOWN BIT(MEM_UNKNOWN) 226#define MEM_FLAG_FPM BIT(MEM_FPM) 227#define MEM_FLAG_EDO BIT(MEM_EDO) 228#define MEM_FLAG_BEDO BIT(MEM_BEDO) 229#define MEM_FLAG_SDR BIT(MEM_SDR) 230#define MEM_FLAG_RDR BIT(MEM_RDR) 231#define MEM_FLAG_DDR BIT(MEM_DDR) 232#define MEM_FLAG_RDDR BIT(MEM_RDDR) 233#define MEM_FLAG_RMBS BIT(MEM_RMBS) 234#define MEM_FLAG_DDR2 BIT(MEM_DDR2) 235#define MEM_FLAG_FB_DDR2 BIT(MEM_FB_DDR2) 236#define MEM_FLAG_RDDR2 BIT(MEM_RDDR2) 237#define MEM_FLAG_XDR BIT(MEM_XDR) 238#define MEM_FLAG_DDR3 BIT(MEM_DDR3) 239#define MEM_FLAG_RDDR3 BIT(MEM_RDDR3) 240 241/** 242 * enum edac-type - Error Detection and Correction capabilities and mode 243 * @EDAC_UNKNOWN: Unknown if ECC is available 244 * @EDAC_NONE: Doesn't support ECC 245 * @EDAC_RESERVED: Reserved ECC type 246 * @EDAC_PARITY: Detects parity errors 247 * @EDAC_EC: Error Checking - no correction 248 * @EDAC_SECDED: Single bit error correction, Double detection 249 * @EDAC_S2ECD2ED: Chipkill x2 devices - do these exist? 250 * @EDAC_S4ECD4ED: Chipkill x4 devices 251 * @EDAC_S8ECD8ED: Chipkill x8 devices 252 * @EDAC_S16ECD16ED: Chipkill x16 devices 253 */ 254enum edac_type { 255 EDAC_UNKNOWN = 0, 256 EDAC_NONE, 257 EDAC_RESERVED, 258 EDAC_PARITY, 259 EDAC_EC, 260 EDAC_SECDED, 261 EDAC_S2ECD2ED, 262 EDAC_S4ECD4ED, 263 EDAC_S8ECD8ED, 264 EDAC_S16ECD16ED, 265}; 266 267#define EDAC_FLAG_UNKNOWN BIT(EDAC_UNKNOWN) 268#define EDAC_FLAG_NONE BIT(EDAC_NONE) 269#define EDAC_FLAG_PARITY BIT(EDAC_PARITY) 270#define EDAC_FLAG_EC BIT(EDAC_EC) 271#define EDAC_FLAG_SECDED BIT(EDAC_SECDED) 272#define EDAC_FLAG_S2ECD2ED BIT(EDAC_S2ECD2ED) 273#define EDAC_FLAG_S4ECD4ED BIT(EDAC_S4ECD4ED) 274#define EDAC_FLAG_S8ECD8ED BIT(EDAC_S8ECD8ED) 275#define EDAC_FLAG_S16ECD16ED BIT(EDAC_S16ECD16ED) 276 277/** 278 * enum scrub_type - scrubbing capabilities 279 * @SCRUB_UNKNOWN Unknown if scrubber is available 280 * @SCRUB_NONE: No scrubber 281 * @SCRUB_SW_PROG: SW progressive (sequential) scrubbing 282 * @SCRUB_SW_SRC: Software scrub only errors 283 * @SCRUB_SW_PROG_SRC: Progressive software scrub from an error 284 * @SCRUB_SW_TUNABLE: Software scrub frequency is tunable 285 * @SCRUB_HW_PROG: HW progressive (sequential) scrubbing 286 * @SCRUB_HW_SRC: Hardware scrub only errors 287 * @SCRUB_HW_PROG_SRC: Progressive hardware scrub from an error 288 * SCRUB_HW_TUNABLE: Hardware scrub frequency is tunable 289 */ 290enum scrub_type { 291 SCRUB_UNKNOWN = 0, 292 SCRUB_NONE, 293 SCRUB_SW_PROG, 294 SCRUB_SW_SRC, 295 SCRUB_SW_PROG_SRC, 296 SCRUB_SW_TUNABLE, 297 SCRUB_HW_PROG, 298 SCRUB_HW_SRC, 299 SCRUB_HW_PROG_SRC, 300 SCRUB_HW_TUNABLE 301}; 302 303#define SCRUB_FLAG_SW_PROG BIT(SCRUB_SW_PROG) 304#define SCRUB_FLAG_SW_SRC BIT(SCRUB_SW_SRC) 305#define SCRUB_FLAG_SW_PROG_SRC BIT(SCRUB_SW_PROG_SRC) 306#define SCRUB_FLAG_SW_TUN BIT(SCRUB_SW_SCRUB_TUNABLE) 307#define SCRUB_FLAG_HW_PROG BIT(SCRUB_HW_PROG) 308#define SCRUB_FLAG_HW_SRC BIT(SCRUB_HW_SRC) 309#define SCRUB_FLAG_HW_PROG_SRC BIT(SCRUB_HW_PROG_SRC) 310#define SCRUB_FLAG_HW_TUN BIT(SCRUB_HW_TUNABLE) 311 312/* FIXME - should have notify capabilities: NMI, LOG, PROC, etc */ 313 314/* EDAC internal operation states */ 315#define OP_ALLOC 0x100 316#define OP_RUNNING_POLL 0x201 317#define OP_RUNNING_INTERRUPT 0x202 318#define OP_RUNNING_POLL_INTR 0x203 319#define OP_OFFLINE 0x300 320 321/* 322 * Concepts used at the EDAC subsystem 323 * 324 * There are several things to be aware of that aren't at all obvious: 325 * 326 * SOCKETS, SOCKET SETS, BANKS, ROWS, CHIP-SELECT ROWS, CHANNELS, etc.. 327 * 328 * These are some of the many terms that are thrown about that don't always 329 * mean what people think they mean (Inconceivable!). In the interest of 330 * creating a common ground for discussion, terms and their definitions 331 * will be established. 332 * 333 * Memory devices: The individual DRAM chips on a memory stick. These 334 * devices commonly output 4 and 8 bits each (x4, x8). 335 * Grouping several of these in parallel provides the 336 * number of bits that the memory controller expects: 337 * typically 72 bits, in order to provide 64 bits + 338 * 8 bits of ECC data. 339 * 340 * Memory Stick: A printed circuit board that aggregates multiple 341 * memory devices in parallel. In general, this is the 342 * Field Replaceable Unit (FRU) which gets replaced, in 343 * the case of excessive errors. Most often it is also 344 * called DIMM (Dual Inline Memory Module). 345 * 346 * Memory Socket: A physical connector on the motherboard that accepts 347 * a single memory stick. Also called as "slot" on several 348 * datasheets. 349 * 350 * Channel: A memory controller channel, responsible to communicate 351 * with a group of DIMMs. Each channel has its own 352 * independent control (command) and data bus, and can 353 * be used independently or grouped with other channels. 354 * 355 * Branch: It is typically the highest hierarchy on a 356 * Fully-Buffered DIMM memory controller. 357 * Typically, it contains two channels. 358 * Two channels at the same branch can be used in single 359 * mode or in lockstep mode. 360 * When lockstep is enabled, the cacheline is doubled, 361 * but it generally brings some performance penalty. 362 * Also, it is generally not possible to point to just one 363 * memory stick when an error occurs, as the error 364 * correction code is calculated using two DIMMs instead 365 * of one. Due to that, it is capable of correcting more 366 * errors than on single mode. 367 * 368 * Single-channel: The data accessed by the memory controller is contained 369 * into one dimm only. E. g. if the data is 64 bits-wide, 370 * the data flows to the CPU using one 64 bits parallel 371 * access. 372 * Typically used with SDR, DDR, DDR2 and DDR3 memories. 373 * FB-DIMM and RAMBUS use a different concept for channel, 374 * so this concept doesn't apply there. 375 * 376 * Double-channel: The data size accessed by the memory controller is 377 * interlaced into two dimms, accessed at the same time. 378 * E. g. if the DIMM is 64 bits-wide (72 bits with ECC), 379 * the data flows to the CPU using a 128 bits parallel 380 * access. 381 * 382 * Chip-select row: This is the name of the DRAM signal used to select the 383 * DRAM ranks to be accessed. Common chip-select rows for 384 * single channel are 64 bits, for dual channel 128 bits. 385 * It may not be visible by the memory controller, as some 386 * DIMM types have a memory buffer that can hide direct 387 * access to it from the Memory Controller. 388 * 389 * Single-Ranked stick: A Single-ranked stick has 1 chip-select row of memory. 390 * Motherboards commonly drive two chip-select pins to 391 * a memory stick. A single-ranked stick, will occupy 392 * only one of those rows. The other will be unused. 393 * 394 * Double-Ranked stick: A double-ranked stick has two chip-select rows which 395 * access different sets of memory devices. The two 396 * rows cannot be accessed concurrently. 397 * 398 * Double-sided stick: DEPRECATED TERM, see Double-Ranked stick. 399 * A double-sided stick has two chip-select rows which 400 * access different sets of memory devices. The two 401 * rows cannot be accessed concurrently. "Double-sided" 402 * is irrespective of the memory devices being mounted 403 * on both sides of the memory stick. 404 * 405 * Socket set: All of the memory sticks that are required for 406 * a single memory access or all of the memory sticks 407 * spanned by a chip-select row. A single socket set 408 * has two chip-select rows and if double-sided sticks 409 * are used these will occupy those chip-select rows. 410 * 411 * Bank: This term is avoided because it is unclear when 412 * needing to distinguish between chip-select rows and 413 * socket sets. 414 * 415 * Controller pages: 416 * 417 * Physical pages: 418 * 419 * Virtual pages: 420 * 421 * 422 * STRUCTURE ORGANIZATION AND CHOICES 423 * 424 * 425 * 426 * PS - I enjoyed writing all that about as much as you enjoyed reading it. 427 */ 428 429/** 430 * enum edac_mc_layer - memory controller hierarchy layer 431 * 432 * @EDAC_MC_LAYER_BRANCH: memory layer is named "branch" 433 * @EDAC_MC_LAYER_CHANNEL: memory layer is named "channel" 434 * @EDAC_MC_LAYER_SLOT: memory layer is named "slot" 435 * @EDAC_MC_LAYER_CHIP_SELECT: memory layer is named "chip select" 436 * @EDAC_MC_LAYER_ALL_MEM: memory layout is unknown. All memory is mapped 437 * as a single memory area. This is used when 438 * retrieving errors from a firmware driven driver. 439 * 440 * This enum is used by the drivers to tell edac_mc_sysfs what name should 441 * be used when describing a memory stick location. 442 */ 443enum edac_mc_layer_type { 444 EDAC_MC_LAYER_BRANCH, 445 EDAC_MC_LAYER_CHANNEL, 446 EDAC_MC_LAYER_SLOT, 447 EDAC_MC_LAYER_CHIP_SELECT, 448 EDAC_MC_LAYER_ALL_MEM, 449}; 450 451/** 452 * struct edac_mc_layer - describes the memory controller hierarchy 453 * @layer: layer type 454 * @size: number of components per layer. For example, 455 * if the channel layer has two channels, size = 2 456 * @is_virt_csrow: This layer is part of the "csrow" when old API 457 * compatibility mode is enabled. Otherwise, it is 458 * a channel 459 */ 460struct edac_mc_layer { 461 enum edac_mc_layer_type type; 462 unsigned size; 463 bool is_virt_csrow; 464}; 465 466/* 467 * Maximum number of layers used by the memory controller to uniquely 468 * identify a single memory stick. 469 * NOTE: Changing this constant requires not only to change the constant 470 * below, but also to change the existing code at the core, as there are 471 * some code there that are optimized for 3 layers. 472 */ 473#define EDAC_MAX_LAYERS 3 474 475/** 476 * EDAC_DIMM_OFF - Macro responsible to get a pointer offset inside a pointer array 477 * for the element given by [layer0,layer1,layer2] position 478 * 479 * @layers: a struct edac_mc_layer array, describing how many elements 480 * were allocated for each layer 481 * @n_layers: Number of layers at the @layers array 482 * @layer0: layer0 position 483 * @layer1: layer1 position. Unused if n_layers < 2 484 * @layer2: layer2 position. Unused if n_layers < 3 485 * 486 * For 1 layer, this macro returns &var[layer0] - &var 487 * For 2 layers, this macro is similar to allocate a bi-dimensional array 488 * and to return "&var[layer0][layer1] - &var" 489 * For 3 layers, this macro is similar to allocate a tri-dimensional array 490 * and to return "&var[layer0][layer1][layer2] - &var" 491 * 492 * A loop could be used here to make it more generic, but, as we only have 493 * 3 layers, this is a little faster. 494 * By design, layers can never be 0 or more than 3. If that ever happens, 495 * a NULL is returned, causing an OOPS during the memory allocation routine, 496 * with would point to the developer that he's doing something wrong. 497 */ 498#define EDAC_DIMM_OFF(layers, nlayers, layer0, layer1, layer2) ({ \ 499 int __i; \ 500 if ((nlayers) == 1) \ 501 __i = layer0; \ 502 else if ((nlayers) == 2) \ 503 __i = (layer1) + ((layers[1]).size * (layer0)); \ 504 else if ((nlayers) == 3) \ 505 __i = (layer2) + ((layers[2]).size * ((layer1) + \ 506 ((layers[1]).size * (layer0)))); \ 507 else \ 508 __i = -EINVAL; \ 509 __i; \ 510}) 511 512/** 513 * EDAC_DIMM_PTR - Macro responsible to get a pointer inside a pointer array 514 * for the element given by [layer0,layer1,layer2] position 515 * 516 * @layers: a struct edac_mc_layer array, describing how many elements 517 * were allocated for each layer 518 * @var: name of the var where we want to get the pointer 519 * (like mci->dimms) 520 * @n_layers: Number of layers at the @layers array 521 * @layer0: layer0 position 522 * @layer1: layer1 position. Unused if n_layers < 2 523 * @layer2: layer2 position. Unused if n_layers < 3 524 * 525 * For 1 layer, this macro returns &var[layer0] 526 * For 2 layers, this macro is similar to allocate a bi-dimensional array 527 * and to return "&var[layer0][layer1]" 528 * For 3 layers, this macro is similar to allocate a tri-dimensional array 529 * and to return "&var[layer0][layer1][layer2]" 530 */ 531#define EDAC_DIMM_PTR(layers, var, nlayers, layer0, layer1, layer2) ({ \ 532 typeof(*var) __p; \ 533 int ___i = EDAC_DIMM_OFF(layers, nlayers, layer0, layer1, layer2); \ 534 if (___i < 0) \ 535 __p = NULL; \ 536 else \ 537 __p = (var)[___i]; \ 538 __p; \ 539}) 540 541struct dimm_info { 542 struct device dev; 543 544 char label[EDAC_MC_LABEL_LEN + 1]; /* DIMM label on motherboard */ 545 546 /* Memory location data */ 547 unsigned location[EDAC_MAX_LAYERS]; 548 549 struct mem_ctl_info *mci; /* the parent */ 550 551 u32 grain; /* granularity of reported error in bytes */ 552 enum dev_type dtype; /* memory device type */ 553 enum mem_type mtype; /* memory dimm type */ 554 enum edac_type edac_mode; /* EDAC mode for this dimm */ 555 556 u32 nr_pages; /* number of pages on this dimm */ 557 558 unsigned csrow, cschannel; /* Points to the old API data */ 559}; 560 561/** 562 * struct rank_info - contains the information for one DIMM rank 563 * 564 * @chan_idx: channel number where the rank is (typically, 0 or 1) 565 * @ce_count: number of correctable errors for this rank 566 * @csrow: A pointer to the chip select row structure (the parent 567 * structure). The location of the rank is given by 568 * the (csrow->csrow_idx, chan_idx) vector. 569 * @dimm: A pointer to the DIMM structure, where the DIMM label 570 * information is stored. 571 * 572 * FIXME: Currently, the EDAC core model will assume one DIMM per rank. 573 * This is a bad assumption, but it makes this patch easier. Later 574 * patches in this series will fix this issue. 575 */ 576struct rank_info { 577 int chan_idx; 578 struct csrow_info *csrow; 579 struct dimm_info *dimm; 580 581 u32 ce_count; /* Correctable Errors for this csrow */ 582}; 583 584struct csrow_info { 585 struct device dev; 586 587 /* Used only by edac_mc_find_csrow_by_page() */ 588 unsigned long first_page; /* first page number in csrow */ 589 unsigned long last_page; /* last page number in csrow */ 590 unsigned long page_mask; /* used for interleaving - 591 * 0UL for non intlv */ 592 593 int csrow_idx; /* the chip-select row */ 594 595 u32 ue_count; /* Uncorrectable Errors for this csrow */ 596 u32 ce_count; /* Correctable Errors for this csrow */ 597 598 struct mem_ctl_info *mci; /* the parent */ 599 600 /* channel information for this csrow */ 601 u32 nr_channels; 602 struct rank_info **channels; 603}; 604 605/* 606 * struct errcount_attribute - used to store the several error counts 607 */ 608struct errcount_attribute_data { 609 int n_layers; 610 int pos[EDAC_MAX_LAYERS]; 611 int layer0, layer1, layer2; 612}; 613 614/** 615 * edac_raw_error_desc - Raw error report structure 616 * @grain: minimum granularity for an error report, in bytes 617 * @error_count: number of errors of the same type 618 * @top_layer: top layer of the error (layer[0]) 619 * @mid_layer: middle layer of the error (layer[1]) 620 * @low_layer: low layer of the error (layer[2]) 621 * @page_frame_number: page where the error happened 622 * @offset_in_page: page offset 623 * @syndrome: syndrome of the error (or 0 if unknown or if 624 * the syndrome is not applicable) 625 * @msg: error message 626 * @location: location of the error 627 * @label: label of the affected DIMM(s) 628 * @other_detail: other driver-specific detail about the error 629 * @enable_per_layer_report: if false, the error affects all layers 630 * (typically, a memory controller error) 631 */ 632struct edac_raw_error_desc { 633 /* 634 * NOTE: everything before grain won't be cleaned by 635 * edac_raw_error_desc_clean() 636 */ 637 char location[LOCATION_SIZE]; 638 char label[(EDAC_MC_LABEL_LEN + 1 + sizeof(OTHER_LABEL)) * EDAC_MAX_LABELS]; 639 long grain; 640 641 /* the vars below and grain will be cleaned on every new error report */ 642 u16 error_count; 643 int top_layer; 644 int mid_layer; 645 int low_layer; 646 unsigned long page_frame_number; 647 unsigned long offset_in_page; 648 unsigned long syndrome; 649 const char *msg; 650 const char *other_detail; 651 bool enable_per_layer_report; 652}; 653 654/* MEMORY controller information structure 655 */ 656struct mem_ctl_info { 657 struct device dev; 658 struct bus_type *bus; 659 660 struct list_head link; /* for global list of mem_ctl_info structs */ 661 662 struct module *owner; /* Module owner of this control struct */ 663 664 unsigned long mtype_cap; /* memory types supported by mc */ 665 unsigned long edac_ctl_cap; /* Mem controller EDAC capabilities */ 666 unsigned long edac_cap; /* configuration capabilities - this is 667 * closely related to edac_ctl_cap. The 668 * difference is that the controller may be 669 * capable of s4ecd4ed which would be listed 670 * in edac_ctl_cap, but if channels aren't 671 * capable of s4ecd4ed then the edac_cap would 672 * not have that capability. 673 */ 674 unsigned long scrub_cap; /* chipset scrub capabilities */ 675 enum scrub_type scrub_mode; /* current scrub mode */ 676 677 /* Translates sdram memory scrub rate given in bytes/sec to the 678 internal representation and configures whatever else needs 679 to be configured. 680 */ 681 int (*set_sdram_scrub_rate) (struct mem_ctl_info * mci, u32 bw); 682 683 /* Get the current sdram memory scrub rate from the internal 684 representation and converts it to the closest matching 685 bandwidth in bytes/sec. 686 */ 687 int (*get_sdram_scrub_rate) (struct mem_ctl_info * mci); 688 689 690 /* pointer to edac checking routine */ 691 void (*edac_check) (struct mem_ctl_info * mci); 692 693 /* 694 * Remaps memory pages: controller pages to physical pages. 695 * For most MC's, this will be NULL. 696 */ 697 /* FIXME - why not send the phys page to begin with? */ 698 unsigned long (*ctl_page_to_phys) (struct mem_ctl_info * mci, 699 unsigned long page); 700 int mc_idx; 701 struct csrow_info **csrows; 702 unsigned nr_csrows, num_cschannel; 703 704 /* 705 * Memory Controller hierarchy 706 * 707 * There are basically two types of memory controller: the ones that 708 * sees memory sticks ("dimms"), and the ones that sees memory ranks. 709 * All old memory controllers enumerate memories per rank, but most 710 * of the recent drivers enumerate memories per DIMM, instead. 711 * When the memory controller is per rank, csbased is true. 712 */ 713 unsigned n_layers; 714 struct edac_mc_layer *layers; 715 bool csbased; 716 717 /* 718 * DIMM info. Will eventually remove the entire csrows_info some day 719 */ 720 unsigned tot_dimms; 721 struct dimm_info **dimms; 722 723 /* 724 * FIXME - what about controllers on other busses? - IDs must be 725 * unique. dev pointer should be sufficiently unique, but 726 * BUS:SLOT.FUNC numbers may not be unique. 727 */ 728 struct device *pdev; 729 const char *mod_name; 730 const char *mod_ver; 731 const char *ctl_name; 732 const char *dev_name; 733 void *pvt_info; 734 unsigned long start_time; /* mci load start time (in jiffies) */ 735 736 /* 737 * drivers shouldn't access those fields directly, as the core 738 * already handles that. 739 */ 740 u32 ce_noinfo_count, ue_noinfo_count; 741 u32 ue_mc, ce_mc; 742 u32 *ce_per_layer[EDAC_MAX_LAYERS], *ue_per_layer[EDAC_MAX_LAYERS]; 743 744 struct completion complete; 745 746 /* Additional top controller level attributes, but specified 747 * by the low level driver. 748 * 749 * Set by the low level driver to provide attributes at the 750 * controller level. 751 * An array of structures, NULL terminated 752 * 753 * If attributes are desired, then set to array of attributes 754 * If no attributes are desired, leave NULL 755 */ 756 const struct mcidev_sysfs_attribute *mc_driver_sysfs_attributes; 757 758 /* work struct for this MC */ 759 struct delayed_work work; 760 761 /* 762 * Used to report an error - by being at the global struct 763 * makes the memory allocated by the EDAC core 764 */ 765 struct edac_raw_error_desc error_desc; 766 767 /* the internal state of this controller instance */ 768 int op_state; 769 770#ifdef CONFIG_EDAC_DEBUG 771 struct dentry *debugfs; 772 u8 fake_inject_layer[EDAC_MAX_LAYERS]; 773 u32 fake_inject_ue; 774 u16 fake_inject_count; 775#endif 776}; 777 778/* 779 * Maximum number of memory controllers in the coherent fabric. 780 */ 781#define EDAC_MAX_MCS 16 782 783#endif 784